Think Your Postgres Backups and Disaster Recovery Are Safe? Let's Talk.
Disaster recovery is perhaps the most ignored aspect of managing PostgreSQL in production systems, and its value is only realized in a moment of crisis. If not done right, absence of backup and recovery strategies can make one repent for days, if not months after. Even when catered to, majority of production systems only have at most one defined backup and recovery strategy, not realizing the benefits of using a hybrid solution or a combination of multiple types of backups and recovery approaches. Contrary to common perception of treating disaster management as a one model fits all philosophy, emphasis should be given on how various types of backups have been optimized for specific kinds of disaster scenarios.
In this talk, I will start off by the most common techniques people backup and restore their data with, and the assumptions, fallacies and disasters they often. The talk will then proceed to the various combinations of backups, monitoring and restores that are ideal for different types of disasters, big and small.
Aside from tools in Postgres core, I will be covering both, popular open source tools (barman, OmniPITR, WAL-e etc.) as well as ways to implement a solid scheme in-house for companies who want to minimize dependence on external tools.
I will also briefly talk about the impact a filesystem can have on the quality of physical backups (ZFS for instance) for audience members who are looking to setup new environments or projects.
Lastly, I will go about automating and provisioning such a setup to minimize maintenance effort and improve simplicity.